Cloud and Data Center
A colocation agreement can run up to a decade, and its power, connection and exit terms decide what the space really costs. We find the right home for each workload, in cloud or colocation, and settle those terms before you sign.
Book a conversationHow Cloud and Colocation Are Bought, and When You Need Help
With public infrastructure as a service, you rent compute, storage and networking, and run your own operating systems and software on top; the provider controls the infrastructure underneath. With colocation, you own the hardware and rent space, power, cooling and connections in someone else's building, usually through one-time charges, a fixed monthly charge for space and power, and usage charges such as remote hands. Many enterprises use both. When private cloud in your own space is joined to public cloud so data and applications can move between them, that is hybrid cloud, and how you design it decides which providers fit.
When outside help makes sense
- A hybrid design, with public cloud alongside private cloud in colocation or your own data center.
73% of organizations use hybrid cloud.
Flexera 2026 State of the Cloud - A move into colocation, where power, cross-connect and exit terms will shape the cost for years.
- A colocation agreement approaching its renewal notice date, with pricing that may reset at renewal.
- A migration that includes mainframe, mid-range or older systems, where one large provider's own guidance calls for careful assessment before any move.
When it does not
In our judgment, you may not need a new provider, or a move, for:
- A single, small cloud account on standard terms. The case grows with the number of sites, providers and agreements.
- A higher Tier than the workload needs. Uptime Institute says a higher Tier is not automatically better. Two sites can protect you more than the rating of one building.
- Workloads with good reasons to stay put, such as data residency rules, unresolved dependencies or a recent hardware refresh. They can stay where they are until those reasons change.
Not every workload should move. Part of our job is telling you which ones stay.
Where Data Center and Cloud Designs Fall Short
The building's rating and the circuit's size are the first numbers buyers see. Neither tells you enough.
Tier ratings and Tier claims
Uptime Institute defines four Tiers. Tier III means any component can be taken out for maintenance without affecting IT; Tier IV adds fault tolerance. Certification comes in stages, and a certificate for the constructed facility is a stronger claim than one for the design documents. "Tier III equivalent" is not a certification. Ask which stage was certified and when, and check Uptime's public directory.
What a Tier does not measure
Uptime assesses operations separately from design, because how a site is run matters alongside how it is built. Its Tiers page attaches no availability percentage to any Tier. The SLA, not the rating, is your protection in the contract.
Circuits you can't load to their rating
Usable power can be less than the circuit rating. Under one provider's terms, you may draw only part of each circuit's rated capacity, and the combined draw on redundant A and B feeds has to fit within one side, so either can carry the whole load if the other fails. Ask for the contractual kW draw cap, per-phase limits and where the meter sits.
Redundant power that counts only when both feeds fail
Under one provider's terms, a redundant power SLA counts an outage only when both feeds to a dual-corded cabinet fail, and only if your equipment fails over on its own. Single-corded equipment gets a lower commitment, and outages while you are over your allocated power can be excluded.
Cooling and density
ASHRAE recommends an inlet air range of 18 to 27 °C (about 64 to 81 °F) for air-cooled IT equipment, as a guideline rather than a rule. Racks for AI, with high power per rack and liquid cooling, are a different purchase, covered under AI and compute.
Cloud availability depends on your design
One large provider's higher compute commitment applies only when instances run across two or more availability zones. Under the shared responsibility model providers publish, patching guest operating systems and setting firewall rules stay with you.
Re-architecting in the middle of a move
One large provider's migration guidance recommends moving first and modernizing afterward, because refactoring during a large migration is the most complex and costly route.
Data that moves between clouds
The UK competition regulator found that the large providers' free-exit programs don't cover ongoing multi-cloud use, so data moving between clouds keeps paying transfer charges. Count that in the design, not after it.
The carriers in the building
Which carriers are present in a colocation site, or restricted from it, decides how you can connect to your network and to the clouds. Cloud connectivity covers the private connections themselves.
Where Each Workload Lands, and What the Terms Cost
For IT and infrastructure leaders: each workload where it runs best
- A decision on which workloads move and which stay, made before any provider is chosen.
- Ready-for-service dates tied to credits and a right to terminate.
- Renewal notices land on TREU's desk, not your infrastructure team's.
For finance: the cost in the definitions
- Power, space and connection charges compared on the same basis across bids.
- Escalators and pass-through power costs known before signature.
- Exit, removal and holdover terms settled at signature, not at the end of the term.
What to Watch For in Colocation and Cloud Contracts
In our judgment, the monthly rate is the easy part. The cost sits in the definitions.
-
Power definitions and overage.
Settle how allocated power is defined, whether you pay for reserved capacity or actual use, what you pay during an outage and what overage costs. Under one provider's terms, overage is billed from the first day it occurred.
-
Escalators and pass-throughs.
Know what the monthly charge includes and how it rises each year, at a fixed rate or tied to an index. Under one provider's terms, prices can rise outside the scheduled review when the provider's own power costs rise.
-
Cross-connects and remote hands.
Recurring connection fees add up over a long term, so price them for the whole term. Under one provider's terms, remote hands work is done at your direction and your risk, with the provider's liability limited to the fee.
-
Ready-for-service dates.
Tie the ready-for-service date to consequences: a provisioning deadline, credits that step up with the delay and a right to terminate if it slips too far.
-
Credits as the only remedy.
Under many SLAs, credits apply to future bills, must be claimed by a deadline, may be capped and are your only remedy unless the agreement says otherwise.
-
Renewals on more than one clock.
Colocation agreements can renew automatically, with strict notice requirements and pricing that resets at renewal. Under one provider's terms, cross-connects and remote hands roll month to month on their own schedule, apart from space and power.
-
Room to grow.
Negotiate the right to add or reduce power and space, and rights to adjacent capacity, before you need them. We favor providers that can grow with you.
-
Liability, casualty and distress.
Liability caps, property damage, insurance and what the provider owes if the building is damaged are among the most heavily negotiated terms. Plan for financial distress on either side: access to your equipment, liens and lien waivers, and deposits or suspension rights tied to objective triggers.
-
Exit and equipment removal.
Settle early termination, removal timelines, data destruction and holdover fees at signature. Under one provider's terms, equipment left past a short removal window is deemed abandoned.
-
Leaving a cloud provider.
The largest cloud providers offer to waive data transfer charges for customers who leave, but on request, for a full exit and for a limited time, and the UK regulator found real limits on those programs. For services in the EU, the Data Act is phasing out switching charges.
-
Proof of controls.
Ask for the provider's SOC 2 report, which covers controls relevant to security, availability, processing integrity, confidentiality or privacy.
This list is a starting point for your review. TREU is not a law firm. Your counsel gives the legal advice.
Colocation and Cloud, Sourced in Four Stages
See how we work-
Supplier identification
We decide with you which workloads belong in cloud, in colocation or where they are, then take each requirement to providers wherever your sites are, regional operators included.
-
Buying cycle
We put power, space and connection charges on the same basis across bids, and negotiate the commercial terms alongside your counsel: power definitions, ready-for-service remedies, growth rights and exit.
-
Implementation oversight and accountability
We hold the provider to the ready-for-service date and the signed terms through installation and acceptance, and check the first invoice against the contract.
-
Client and supplier lifecycle management
We track every notice date, including services that renew on their own schedule, and join business reviews. In our experience, the work starts months before the notice date, not weeks. We renegotiate each renewal and source new needs for the life of the agreement.
Questions Buyers Ask About Cloud and Colocation
What does a data center Tier rating actually tell us?
It describes how the site's infrastructure is designed and built: a Tier III site can take any component out for maintenance without affecting IT, and a Tier IV site is fault tolerant. It doesn't certify how well the site is run, and Uptime itself says a higher Tier is not automatically better.
Is a Tier III design certification the same as a Tier III facility?
No. Design certification shows the drawings meet the Tier, and Uptime treats it as a provisional step that expires two years after award. Certification of the constructed facility is the stronger claim. "Built to Tier III" is not a certification at all.
How is colocation power priced?
Common models include a flat fee per circuit, a committed kW allocation, metered use with a minimum charge, and an all-in rate per kW that bundles space and power. Metered models often add a facility overhead factor, so make sure the contract defines how that factor is calculated.
Why can't we load our circuits to their full rating?
Usable power can be less than the circuit rating. Under one provider's terms, you may draw only part of each circuit's rated capacity, and with A and B feeds, each side must be able to carry the whole load alone if the other fails.
What happens to our equipment if we leave, or the provider fails?
That depends on the exit and distress terms you negotiate: removal timelines, data destruction, access to your equipment and lien waivers. Under one provider's terms, improvements to the space become the provider's property at termination, while movable hardware stays yours.
What does it cost to leave a cloud provider?
The largest providers offer to waive data transfer charges for a full exit, on request and for a limited time. Partial moves and ongoing multi-cloud use are generally not covered, so plan the exit terms before you need them.
Should we move everything to the cloud?
In our judgment, rarely. Hybrid is the common pattern, and some workloads have good reasons to stay, such as data residency rules or a recent hardware refresh. Where you do move, one large provider's own guidance is to move first and modernize afterward.
What stays our responsibility in IaaS?
Under the shared responsibility model providers publish, you remain responsible for the guest operating system and its patches, your applications and the configuration of firewall controls. The provider runs the hardware, the virtualization layer and the buildings.